Related Experiment Video
Updated: Mar 7, 2026
07:38
Microfluidic Fabrication of Polymeric and Biohybrid Fibers with Predesigned Size and Shape
Published on: January 8, 2014
9.3K
FibreCastML: an open web platform for predicting electrospun nanofibre diameter distributions for biomedical
Elisa Roldán1, Kirstie Andrews1, Stephen M Richardson2
1Department of Engineering, Faculty of Science & Engineering, Manchester Metropolitan University, Manchester, United Kingdom.
Frontiers in Bioengineering and Biotechnology
|March 6, 2026
Summary
FibreCastML is a new machine learning framework that predicts the full fiber diameter distribution in electrospinning, moving beyond mean-only predictions. This enables more accurate and reproducible scaffold design for tissue engineering and drug delivery applications.
Area of Science:
- Biomaterials Engineering
- Nanotechnology
- Computational Science
Background:
- Electrospinning generates tunable fibrous scaffolds for biomedical applications.
- Current machine learning models predict only mean fiber diameters, neglecting crucial distribution information.
- Scaffold functionality and biomimicry depend on the entire fiber diameter distribution.
Purpose of the Study:
- Introduce FibreCastML, an open-access, distribution-aware machine learning framework.
- Predict full fiber diameter spectra from standard electrospinning parameters.
- Provide interpretable insights into how processing parameters influence fiber diameter distribution.
Main Methods:
- Curated a meta-dataset of 68,538 fiber diameter measurements from 1,778 studies.
- Trained seven machine learning learners using six standard input parameters.
- Employed nested cross-validation with leave-one-study-out external folds for generalizable performance.
- Integrated variable importance, SHapley Additive exPlanations (SHAP), and 3D parameter maps for interpretability.
Main Results:
- Non-linear learners achieved R² > 0.91 for multiple biomedical polymers.
- Solution concentration was identified as the most influential global variable.
- FibreCastML accurately predicted fiber diameter distributions, validated experimentally (Kolmogorov-Smirnov p > 0.13).
Conclusions:
- FibreCastML establishes a new paradigm for electrospinning design by focusing on distribution-aware modeling.
- The framework enables reproducible, sustainable, and data-driven optimization of scaffold architecture.
- FibreCastML empowers global research for faster, greener, and more reproducible electrospinning, advancing nanomanufacturing and biomedical innovation.
